Synthesis and characterization of materials. BNNF-Cu with different copper contents were prepared via a facile method using low-cost raw materials. CuCl2, melamine and boric acid were first reacted to form a gel-like BNNF-Cu precursor. BNNF-Cu was obtained by annealing the precursor at 900 oC as stated in detail in the Methods section. Catalytic activity of the BNNF-Cu was optimized by regulating the amount of copper salt added. As shown in Supplementary Fig. 1–2, the BNNF-Cu samples prepared with different copper concentrations have similar crystal structures and functional groups. Supplementary Fig. 3 shows that BNNF-Cu-2 has the highest catalytic activity. Thus, subsequent studies would be concentrated on BNNF-Cu-2 and it will be simply referred as BNNF-Cu hereafter. Figure 1a shows a schematic diagram of BNNF-Cu’s preparation process.
Microstructure and morphology of BNNF-Cu were characterized with field-emission scanning electron microscopy (FE-SEM) and transmission electron microscopy (TEM). As shown in Fig. 1b-d, the as-prepared sample showed a fiber-like structure with a rectangular cross-section (see Supplementary Fig. 4) of about 200 nm width and relatively low crystallinity (Fig. 1c, inset). There is no obvious difference between BNNF-Cu and BNNF synthesized without Cu (Supplementary Fig. 5), indicating that the addition of Cu2+ ions does not affect the samples’ morphology. Fourier transform infra-red (FT-IR) spectra of BNNF and BNNF-Cu were shown in Supplementary Fig. 6. A spectrum from commercially purchased boron nitride nanosheets (BNNS) was also shown for comparison. All the three samples clearly show two characteristic B-N vibrations at ∼1380 cm− 1 (B − N transverse stretching in plane) and ∼800 cm− 1 (B − N−B bending out of plane). There are also additional peaks in BNNF and BNNF-Cu located between 3200 ∼ 3600 cm− 1 relating to functional groups -OH and ν (N − H) and a weak vibration at 1637 cm− 1 probably due to δ (N − H)28. The introduction of these functional groups probably leads to enhancement of adsorption capacity. The counter-phase B − N E2g Raman vibration mode of boron nitride at ∼1370 cm− 1 can be observed in Raman spectra of all three samples (Supplementary Fig. 7)38. The main difference is that the Raman peak is much shaper in BNNS which is highly crystalline hexagonal boron nitride (Fig. 1g and Supplementary Fig. 8). The noisy spectra, board and shifted Raman peaks of BNNF and BNNF-Cu are due to their low crystallinity. Supplementary Fig. 9 shows X-ray photoelectron spectroscopy (XPS) survey spectra from BNNF and BNNF-Cu. Other than B and N, signal from O and C are attributed to surface contamination. It is interesting to note that almost no Cu signal was detected even in the BNNF-Cu sample, suggesting that the amount of Cu in it is close to or lower than detection limit of our XPS system. The above analysis shows that adding Cu2+ in the reactant will not cause significant impact on the morphology, crystal structure nor functional groups of the formed BNNF or BNNF-Cu. We can also conclude that the amount of Cu in the BNNF-Cu should be low (< 1%).
We then employed more sophisticated analytical techniques including synchrotron radiation analysis, high-resolution high-angle annular dark-field (HAADF) transmission electron microscopy as well as electron paramagnetic resonance (EPR) spectroscopy to determine whether there is really copper in the BNNF-Cu sample as well as the distribution and nature of copper if it does exist. Elemental HAADF mapping of B, N, Cu in BNNF-Cu (Fig. 1e and Supplementary Fig. 10) verified the homogeneous distribution of Cu. Importantly, many blue single spots in the element mapping image of partial BNNF-Cu and bright spots in the high-resolution high-angle annular dark-field (HAADF) image (Fig. 1f) both indicated that isolated (or few-atom clusters of) Cu atoms were successfully and evenly anchored on BNNF. X-ray diffraction (XRD) patterns of BNNF, BNNF-Cu and BNNS show two primary diffraction peaks at about 26.5° and 42.5°, corresponding to the (002) plane and (100)/(101) of hexagonal BN (PDF34-0421), respectively30. Interestingly, comparing the peaks of these three layered samples, the (002) peak of BNNF-Cu is slightly shift to lower angel, meaning that the interlayer distance of BNNF-Cu is expanded probably attributed to the insertion of Cu atoms between BN layers40. Furthermore, the decrease of the primary diffraction peak intensity for BNNF-Cu, in contrast to BNNF, suggests that the insertion of Cu atoms does distort the crystal structure of BNNF to some extent41. In addition, the specific surface area of BNNF-Cu was determined to be 153 m2 g− 1 which is higher than that of BNNS and BNNF (Supplementary Fig. 11 and Supplementary Table 1). The improvement of specific surface area can be attributed to the existence of Cu atoms providing more adsorption sites and increasing the probability of pore formation.
To further investigate the dispersion of Cu atoms and coordination between Cu atoms and BNNF, synchrotron radiation X-ray absorption near-edge spectroscopy (XANES) and extended X-ray absorption fine structure (EXAFS) measurement were conducted. Three standard samples of copper foil, CuO and Cu2O were used as references. The obtained results reveal that the Cu species are dispersed as isolated single Cu atoms and stabilized by nitrogen atoms of BNNF. As shown in Fig. 2a, the absorption edge for BNNF-Cu is sandwiched between those of CuO and Cu2O, clearly demonstrating the electron structure with positive charges in Cu δ+ (1 < δ < 2)40. In addition, an electron paramagnetic resonance (EPR) spectrum of BNNF-Cu shows the characteristic Cu2+ signal (g∥=2.33, g⊥=2.05)42 (Fig. 2d) and a XPS high resolution Cu 2p spectrum of BNNF-Cu (Supplementary Fig. 12) verified the existence of Cu2+ in BNNF-Cu40. Figure 2b shows Fourier transform of the Cu K-edge EXAFS of BNNF-Cu and reference samples Cu foil, Cu2O and CuO. There is a main peak at ~ 1.6 Å of BNNF-Cu, which is closing to the main peak positions of CuO and Cu2O, can be attributed to the Cu-N first shell. Additionally, no obvious peak at 2.2 Å (metallic Cu-Cu coordination) can be observed. These results reveal that the atomic dispersion of Cu atoms in BNNF-Cu. Moreover, the wavelet transforms (WT) of Cu K-edge EXAFS also present the well dispersion of Cu atoms in BNNF-Cu (Fig. 2c). The WT for the EXAFS signals of BNNF-Cu, Cu foil, CuO2, and CuO are achieved by using the complex wavelet developed by H. Hunke43. The WT contour plots of BNNF-Cu shows one maximum intensity at point (4 Å−1, 1.6 Å), which can be ascribed to the Cu-N coordination, and almost no Cu-Cu signals are observed, as compared with that of Cu foil, CuO, and CuO2. Structural parameters of BNNF-Cu were further extracted by fitting the Cu K-edge EXAFS spectrum using the Artemis software. Figure 2e and Supplementary Fig. 13 show optimal fitting curves in R space and K space, respectively. The optimal fitting results are attributed to the first shell of Cu-N (bond distance: ~ 2.0 Å) and the second shell of Cu-B (bond distance: ~ 2.64 Å), as shown in Supplementary Tables 2 and 3. Moreover, there is a significant difference in the coordination number (CN) between the first shell (Cu-N: 6.26 ± 0.97) and the of second shell (Cu-B: 2.56 ± 0.74), indicating that the coordination structure around the Cu single atoms is very disordered, which is consistent with XRD results. According to these results, we then build a model for density functional theory (DFT) calculation with an isolated Cu atomic surrounded with six Cu-N as shown in Supplementary Fig. 14.
Photocatalytic activity of BNNF-Cu. Aromatic amines are important intermediates with wild usages for synthesis of many chemicals for commercial and biomedical applications. Selective catalytic reduction of nitro compounds into amines with hydrogen or hydrogen donors has been considered as an important chemical reaction in synthetic organic chemistry44. Herein, we selected the reduction of p-nitrophenol (PNP) to p-aminophenol (PAP) with borohydride to evaluate the catalytic performance of BNNF-Cu. It is known that after adding NaBH4 to a PNP solution, 4-nitrophenolate anions with strong visible absorbance at 400 nm will be formed (Supplementary Fig. 15), and the concentration of 4-nitrophenolate is proportional to the intensity of the absorption peak. Under irradiation, the characteristic absorption peak of PNP gradually disappears, accompanied by the appearance PAP absorption peak centered at about 300 nm during the catalytic reduction of PNP over BNNF-Cu (Fig. 3a). Moreover, the existence of two isosbestic points at 280 and 314 nm indicates complete conversion of PNP to PAP without any byproducts45. As shown in Fig. 3b, no matter under irradiation or in dark, the Ct/C0 (where Ct is the concentration of PNP at time t) value did not change in BNNS and BNNF, indicating that these two materials cannot catalyze the reduction of PNP. By contrast, the addition of BNNF-Cu caused a rapid decrease of PNP concentration whether under photoexcitation (BNNF-Cu-D-L) or in dark (BNNF-Cu-D). These results suggested that the present of copper is essential for the catalytic action of BNNF-Cu. To further confirm this point, we decorated a BNNF sample with copper nanoparticles (BNNF-Cu-NP) with sonicating a BNNF aqueous dispersion with pre-prepared copper nanoparticle of 50 to 80 nm diameter (Supplementary Fig. 16). Figure 3b shows that the surface decoration of copper nanoparticles onto BNNF indeed endow it with catalytic activity both under photoexcitation and in dark. On the other hand, it can also be noted that catalytic activity of the BNNF-Cu sample is much higher than that of BNNF-Cu-NP suggesting that the isolated copper atoms or few-atom clusters on BNN-Cu do have better catalytic performance comparing to copper nanoparticles. Since the initial amount of the NaBH4 added is far more than needed, a pseudo first-order kinetic equation has been utilized to evaluate reaction rate, written as ln (Ct/C0) = -kt, (where k is the apparent rate constant). The calculated rate constant (k) of BNNF-Cu-L (0.65 min− 1) is more than twice that of BNNF-Cu-D (0.31 min− 1) (Supplementary Fig. 17a, c), indicating a good photocatalytic activity of the BNNF-Cu. To clarify the influences of visible light excitation on the catalytic performance, a 400 nm filter has been equipped on Xenon lamp to block ultraviolet (UV) light with wavelength below 400 nm. As shown in Fig. 3c, without UV light, the BNNF-Cu performed a moderate catalytic activity (k ≈ 0.43, Supplementary Fig. 17b), indicating that the BNNF-Cu has a certain sensitivity to visible light.
To check the stability of the catalyst, the PNP reduction experiment was repeated by adding the same amount of PNP into the used solution after complete disappearance of original PNP’s absorption peak. After 5 cycles, photocatalytic activity of BNNF-Cu shows little observable degrade (Fig. 3d). Compared the XRD pattern and FTIR spectrum of BNNF-Cu after reaction with that of the as-prepared sample, the almost same results (Supplementary Fig. 18) indicates that the crystallinity and composition of the catalyst did not change during reaction. Meanwhile, there is no observable aggregation of the isolated Cu atoms in BNNF-Cu after use (Supplementary Fig. 19). Thus, the BNNF-Cu shows a high operation stability under photoexcitation. Moreover, the photocatalytic universality of BNNF-Cu also estimated by reducing other nitroaromatic chemicals. As shown in Fig. 3e, f, the BNNF-Cu can quickly convert o-nitrophenol, 3-nitrophenol, 3-methyl-4-nitrophenol, 2-chloro-4-nitrophenol, and 2-methyl-4-nitrophenol to their corresponding reductive amino-aromatic products under irradiation. To show the excellent performance of BNNF-Cu, a comparison of the catalytic activities with that of the recently reported catalysts was made (Fig. 3g). Because the mass of metal in catalysts and the amount of PNP in solution varied largely in different reported works, the normalized turnover frequency (TOF) was employed to evaluate the catalytic activity. As shown in Fig. 3, BNNF-Cu shows a record high TOF comparing to previously reported non-precious-metal-based catalysts. (left panel of Fig. 3g) and even higher than many precious-metal-based catalysts (right panel of Fig. 3g). Although the TOF of BNNF-Cu is still lower than a few precious-metal-based catalysts, considering the price of copper which is much lower than precious metal (Au, Ag, Pt and Pd) (Cost per TOF are shown in Supplementary Table 4), the BNNF-Cu does possess higher cost-efficiency. All the above-mentioned excellent performance indicated that the BNNF-Cu catalyst has potential application prospects in the field of nitro reduction synthesis of amines.
Mechanism of enhanced catalytic performance. To further explain the positive effect of Cu single atoms in BNNF-Cu on catalytic activity, we systematically analyzed the band structure and electrical conductivity of BNNF-Cu. Compared with BNNF, the BNNF-Cu sample have a narrower bandgap, higher valence band position and broader light absorption range (Supplementary Fig. 20–22), revealing that the introduction of a small amount of Cu single atoms can change the band positions and improve light absorption capacity. Figure 4a shows the time-resolved photoluminescence (TRPL) spectroscopy of BNNF and BNNF-Cu. The PL decay lifetime (τ) of BNNF-Cu (2.40 ns) is slightly shorter than that of BNNF (2.73 ns), indicating an enhancement of charge transfer and separation in the BNNF-Cu system, which are probably ascribed to narrower bandgap and new transfer channels caused by anchoring of isolated Cu atoms10,40,46. Higher photocurrent density (Fig. 4b) and lower impedance of BNNF-Cu (Fig. 4c) also support the conclusion that the efficiencies of charge separation and transfer are improved by the induced isolated Cu atoms. As a result, anchoring small amount of isolated Cu atoms on BNNF can simultaneously adjust band structure and enhance the separation and transfer of electron-hole pairs.
DFT calculations were applied to gain a deeper understanding of the excellent catalytic performance of BNNF-Cu in the reduction of PNP to PAP. According to some previous works and our calculations, the configurations of adsorbate on BNNF-Cu were set to be parallel to the catalyst surface47,48. Figure 5 shows schematic diagram of the distal pathway for reduction of PNP to PAP catalyzed by BNNF-Cu and comparison of the free-energy diagrams of the reduction of PNP. In detail, after PNP adsorbed on BNNF-Cu (ΔE = − 1.19 eV), the first hydrogenation reaction happens that the oxygen atom of PNP combines with a hydrogen atom to form a *NOOH intermediate (ΔE = − 0.89 eV). Subsequently, the *NOOH is dehydroxylated to *NO (ΔE = − 0.20 eV), followed by a second hydrogeneration that the H atom connected the N atom or O atom forming *NHO (ΔE = − 0.85 eV) or *NOH (ΔE = − 0.70 eV) intermediates, separately. After that, a hydrolysis reaction occurs and a new intermediate (*NHOH, ΔE = − 0.82 eV or − 0.97 eV) is formed, which is caused by the *NHO or *NOH groups with strong Brønsted acidity. Then, the *NHOH is dehydroxylated to *NH (ΔE = − 1.16 eV) before the formation of the final product of PAP (ΔE = − 1.23 eV) by the third hydrogenation. The hydrogenation reduction of PNP to PAP follows the reaction pathway that PNP (*NO2) →*NOOH →*NO →*NOH (*NHO) →*NHOH →*NH →PAP (*NH2) in our BNNF-Cu system. There is no energy barrier for the hydrogenation of PNP for BNNF-Cu in the reaction pathway, in contrast, there is a large energy barrier for PNP (*NO2) →*NOOH for BNNF (Supplementary Fig. 23), indicating that anchoring isolated Cu atom on BNNF makes it suitable for reduction reactions.